A mechanical system is entirely described by the single mathematical model coupling the system's design with mechanical laws; given the system's initial state and the control strategy, the model predicts its future trajectory precisely. Looking at a complex socio-environmental system, we find ourselves in a different situation. The laws driving the system are poorly understood, and the system is subject to uncertain inputs. Any particular model captures the system only partially, and, hence, the use of a single model can be misleading. We suggest using a multi-model approach aiming at integrating knowledge given by several models showing the system from different angles. In this presentation two possible approaches to integration of models are discussed. The first one shows how to integrate alternative random variables describing the same observed quantity based on assessment of the joint probability distribution conditioned to the posterior event. The second approach allows for generating a synthetic trajectory from two alternative model-based trajectories approximating the actual past dynamics in an appropriate metrics.